The licence that is not there

Every one of the twenty-one op-eds says you may republish it. Not one of them says so under a licence. This page is the difference between those two sentences, counted in the bytes.

What is actually on the pieces

At the foot of every article, after the prose and before the author’s role line, there is one sentence. It is the only statement of terms anywhere on the page. It appears in two wordings across the corpus, which differ in the occasion they name:

Quoted in full, twice, because this is the one text on which every other claim on this page depends, and a reader cannot check our reading of it against anything else. Both wordings are in the frozen bytes verbatim; the gate fails the build if they stop matching.

The count

Each number below is derived from the frozen HTML by build/graph.py and then derived again, by a different program with different patterns, in build/gates.py. A number that only one program can produce is an assertion; a number two programs produce from the same bytes is a finding.

What that sentence does and does not do

• It permits republication and translation IN FULL. It does not permit an extract.
• It requires translations to remain faithful to the original meaning, which is a condition no standard public licence states in those words.
• It permits minor edits for length or house style, which sits oddly beside 'in full'.
• It names no licence, no version, no URL and no identifier, so there is nothing to cite, point at, or check against.
• It says nothing about attribution. The announcement page says 'with appropriate credit'; the articles do not mention credit at all.
• It says nothing about commercial use, derivative works, moral rights, term, or revocation.
• It is not addressed: it does not say who grants it, and the pieces carry no copyright notice naming a holder.

Two statements, one corpus

The same organisation states the terms twice, in two places, and the two do not say the same thing.

Both sentences are now read from frozen, hashed bytes. The announcement’s was not, until v0.4.2: wan-ifra.org sits behind a JavaScript-challenge firewall that refused two automated attempts, and this section published that as “the one page in the beat a machine cannot read”, which was too strong. The refusal is intermittent; a later attempt returned the page, and the sentence quoted by hand turned out to be verbatim. What we got wrong, and the rule it produced →

The machine-readable form that is nearly there

This is the part that is genuinely frustrating, because almost all of the work is already done.

Every one of the twenty-one pages already ships schema.org structured data as JSON-LD, generated automatically by the site’s SEO plugin. The types present across the corpus are Article, BreadcrumbList, ImageObject, Person, WebPage, WebSite. schema.org has carried a license property since 2015, alongside license, copyrightHolder, copyrightNotice, usageInfo, isAccessibleForFree. 0 of 21 use any of them.

The record is being generated on every page already. Adding the terms to it is a line:

// in the Article node that is already on every one of the 21 pages
"license": "https://creativecommons.org/licenses/by/4.0/",
"copyrightHolder": { "@type": "Organization", "name": "World News Day" },
"isAccessibleForFree": true

And the structured data that is there is not reliable. The REST record for each piece carries an author field. On 2 of the 21 it does not name the people the page itself prints: one gives the CMS account name 329976pwpadmin instead of a person, and one names an author the byline does not. A machine-readable field that is wrong twice in twenty-one is worse than no field at all, because nothing on the page tells a reader which to believe. It is why the bylines in the corpus are read from the prose.

There is more that is nearly there. worldnewsday.org serves the entire corpus as structured JSON at /wp-json/wp/v2/posts — every piece, with its title, author field, dates, categories and full body, in a form any program can read. It is open, undocumented and unadvertised: a machine-consumable edition exists by accident of the platform rather than by a decision to publish one. It carries no licence field either. And the taxonomy is a single category (Story) and a single tag (2026) across all twenty-one, which is why the themes in the aggregate had to be derived here rather than read from the source.

What would close the gap

1. Name the licence. “Republish or translate in full, faithful to the original meaning” is close to CC BY-ND 4.0 with a translation permission, and “with appropriate credit” is CC BY 4.0. Either is a URL, a version and a body of case law instead of a sentence to interpret. Publishing the wrong-shaped named licence is still better than publishing none, and the licences are designed to be read by the people this corpus is aimed at.
2. Put it in the JSON-LD. One license line in a record that is already generated makes twenty thousand words of deliberately open material discoverable, indexable and usable as open material, by every machine that reads the page.
3. Say it once. One canonical statement of the terms, linked from every piece and from the announcement, instead of two wordings on the articles and a third on the page that points to them.

What this page is not saying

It is not saying the permission is fake, grudging or unusable. It is the opposite of those things: twenty-one commissioned pieces, given away, in a year when almost nobody in this industry gives anything away. The generosity is the story. The finding is about the form, not the existence — and the form is what decides whether the generosity reaches a translator in Manila, a local paper in Ohio, an archive in Lisbon or a model that will summarise all twenty-one for somebody who will never see the originals.

It is also not saying that this is unusual. It is completely usual. That is why it is worth counting: the gap between “we want this to spread” and “here is the machine-readable statement that lets it spread” is where most of the open content on the internet currently falls, and closing it costs an afternoon.